University of Groningen

Modelling toluene oxidation

Hoorn, J.A.A.; van Soolingen, J.; Versteeg, Geert

Published in:
Chemical Engineering Research and Design

DOI:
10.1205/cherd.04161

IMPORTANT NOTE: You are advised to consult the publisher's version (publisher's PDF) if you wish to cite from
it. Please check the document version below.

Document Version
Publisher's PDF, also known as Version of record

Publication date:
2005

Link to publication in University of Groningen/UMCG research database

Citation for published version (APA):

Hoorn, J. A. A., van Soolingen, J., & Versteeg, G. F. (2005). Modelling toluene oxidation: Incorporation of
mass transfer phenomena. Chemical Engineering Research and Design, 83(2), 187-195. DOI:
10.1205/cherd.04161

Copyright
Other than for strictly personal use, it is not permitted to download or to forward/distribute the text or part of it without the consent of the
author(s) and/or copyright holder(s), unless the work is under an open content license (like Creative Commons).

Take-down policy
If you believe that this document breaches copyright please contact us providing details, and we will remove access to the work immediately
and investigate your claim.

Downloaded from the University of Groningen/UMCG research database (Pure): http://www.rug.nl/research/portal. For technical reasons the
number of authors shown on this cover page is limited to 10 maximum.

Download date: 05-03-2018

 0263–8762/05/$30.00+0.00
# 2005 Institution of Chemical Engineers
www.icheme.org/journals Trans IChemE, Part A, February 2005
doi: 10.1205/cherd.04161 Chemical Engineering Research and Design, 83(A2): 187–195

MODELLING TOLUENE OXIDATION

Incorporation of Mass Transfer Phenomena
J. A. A. HOORN1
, J. VAN SOOLINGEN1 and G. F. VERSTEEG2
1
DSM Research, Geleen, Netherlands
2
Department of Chemical Engineering, University of Twente, Enschede, Netherlands

T
he kinetics of the oxidation of toluene have been studied in close interaction with the
gas –liquid mass transfer occurring in the reactor. Kinetic parameters for a simple
model have been estimated on basis of experimental observations performed under
industrial conditions. The conclusions for the mass transfer and reaction regime on basis of
experimental observations and model calculations are in good agreement: toluene oxidation
under industrial conditions can be characterized as a slow reaction with respect to mass
transfer.

Keywords: chemical reactors; kinetics; mass transfer; mathematical modelling; multiphase

reactions; reaction engineering.

INTRODUCTION KINETICS AND MASS TRANSFER

The oxidation of toluene with air at elevated temp-
The site of DSM Special Products Rotterdam has the
eratures and pressures proceeds via a mechanism consisting
largest toluene oxidation plant in the world, producing ben-
of a chain of radical reactions and is initialized by cobalt
zoic acid that for the major part is converted to phenol. In
ions. The use of elementary reactions in this radical mech-
addition, benzoic acid and its derivative sodium benzoate
anism is at this moment too complicated (lack of basic data
also find their applications in the food and fine chemicals
and the need for extensive numerical solving techniques) to
industry. The DSM site at Rotterdam also has the largest
be of much practical use in the analyses on mass transfer
production unit for benzaldehyde, a specialty chemical.
and reaction. Instead a more simplified kinetic description
Since the development of the process (Keading et al.,
is applied in the present study. The kinetic scheme in
1965) a large number of studies on the chemistry of toluene
Figure 1 is directly derived from the stoichiometric
oxidation and related hydrocarbons have been published.
equations (assuming elementary reactions) and has been
Most of the earlier publications have in common that the
proposed earlier for toluene oxidation (Quiroga et al.,
main focus is on the kinetics and that other process items
1980) as well as xylene oxidation (Cao et al., 1994). For
are not considered (Borgaonkar et al., 1984; Lozar et al.,
oxidation reactions proceeding through radical intermedi-
2001; Mulkay and Rouchaud, 1967; Quiroga et al., 1980;
ates, a zero-order reaction rate is commonly encountered
Scott and Chester, 1972). This approach is not sufficient
(Helfferich, 2001). Morimoto and Ogata (1967) concluded
when considering scale-up and modelling of the plant reac-
from their batch experiments that for oxygen levels below
tors. For safety reasons, oxygen levels in the reactor off-gas
42 vol% (the oxygen concentration in the reactor was
have to be kept below the explosion limits. This means that
kept constant by maintaining a constant pressure through
at least in part of the reactor the oxygen partial pressure is
suppletion of pure oxygen to the oxygen/nitrogen mixture)
so low that the liquid concentration is significantly
the reaction order in oxygen can be taken as one, whereas at
decreased. In the range of low oxygen concentrations, the
higher oxygen levels the order is zero. Bhattacharya et al.
radical chain mechanism produces different by-products
(1973) give a reaction order for oxygen of 12, but the exper-
(Bateman, 1951). In addition, it is expected that the mass
imental conditions are not fully given. The rates for
transfer of oxygen influences the productivity of the reac-
the reactions in Figure 1 are assumed to be first-order in
tor. It is essential to quantify kinetics in close interaction
the reactants, also for oxygen. As it mathematically can
with the mass transfer phenomena occurring in the reactor
be shown, it is not possible to determine the individual
in order to increase the control of the productivity and
rate constants for reactions 1, 2 and 4 independently with-
selectivity.
out experiments comprising benzyl alcohol added to the


Correspondence to: J. A. A. Hoorn, DSM Research, CT&A-ACES,
feed. These experiments have not been performed while
PO Box 18, 6160 MD Geleen, The Netherlands. the availability of the experimental set-up for kinetic exper-
E-mail: johan.hoorn@dsm.com iments was very limited. In addition, benzyl alcohol is

187
188 HOORN et al.

Figure 1. Basic scheme for the kinetics.

known to inhibit the oxidation (Morimoto and Ogata, 1967) kinetic scheme that has a close resemblance to the toluene
making the operation of a continuous tank reactor more oxidation scheme given in Figure 1. The mass transfer
difficult than it is already. A correlation between R1 and flux for a first-order irreversible reaction according to
R2 was made on basis of the observation that in the the Danckwerts modification of the penetration theory is
oxidation mechanism the primary termination reaction defined by (Westerterp et al., 1984):
takes place between two alkoxy radicals (Russell, 1957;   pffiffiffiffiffiffiffiffiffiffiffiffiffiffi
Kondratiev, 1969) whereby alcohol and aldehyde are cA,b
formed in a 1 : 1 ratio. For this reason, the parameters k1 J ¼ kL  cA,i   1 þ w2 (6)
1 þ w2
and k2 were replaced by a single parameter k12 giving the
following expressions for the reaction rates: The extremes for small and large Hatta numbers are
convenient for simple evaluations on basis of analytical
R1 ¼ k12  cTOL  cO2 (1) solutions.
R2 ¼ k12  cTOL  cO2 (2)
w . 2 J ¼ kL  cA,i  w (7)
R3 ¼ k3  cBALD  cO2 (3)
w , 0:3 J ¼ kL  (cA,i  cA,b ) (8)
R4 ¼ k4  cBALC  cO2 (4)
The enhancement factor is the ratio of the flux in pre-
The influence of the catalyst was not included in the rate sence of reaction [equation (6)] and pure physical mass
equations because all experimental data have been per- transfer under identical conditions [expression is equivalent
formed at a constant cobalt concentration. The expression to equation (8)].
for the Hatta number (w) depends on the reaction rate
 
equation. For a first-order irreversible reaction with respect pffiffiffiffiffiffiffiffiffiffiffiffiffiffi cA,i  (cA,b )=(1 þ w2 )
to the absorbing component w is defined by: E ¼ 1 þ w2  (9)
cA,i  cA,b
pffiffiffiffiffiffiffiffiffiffi
kD The extremes for small and large Hatta numbers are:
w¼ (5)
kL
w.2 E¼w (10)
For systems consisting of multiple reactions involving w , 0:3 E ¼ 1 (11)
the gaseous component the most commonly applied simpli-
fication is to evaluate the Hatta number on the assumption The Hatta number and the enhancement factor are cal-
that an overall pseudo zero-order or first-order rate equation culated from experimental observations, respectively
is applicable. The model of the xylene oxidation by Cao included in the modelling.
et al. (1994) comprises a zero order reaction for oxygen
where Hatta is calculated according to the correlation of
Hikita and Asai (1964). This approach is also applied by
EXPERIMENTS AND OBSERVATIONS
Suresh et al. (1988) in the modelling of the cyclohexane
oxidation. Pohorecki et al. (2001) apply a first order depen- The reactor and auxiliary equipment for the toluene
dence for oxygen in the cyclohexane oxidation with a oxidation experiments are operated in a continuous mode,

Trans IChemE, Part A, Chemical Engineering Research and Design, 2005, 83(A2): 187–195
 MODELLING TOLUENE OXIDATION 189

a schematic drawing of the set-up is given in Figure 2. All oxygen is neglected. The experimentally observed oxygen
equipment is accommodated in an explosion proof facility. transfer rate FGL is related with the flux through the inter-
From the continuous monitoring of the reactor temperature facial area:
and the oxygen concentration in the off-gas during exper-
iments it appeared that small fluctuations are present in FGL ¼ J  VR  a (12)
the reaction set-up. The standard deviations of the results
of five reference experiments having identical process con- In case w . 2 substitution of the expression for the
ditions are shown in Table 1 as an indication of the extent observed transfer rate gives (taking into account the defi-
of experimental inaccuracies. The range of process vari- nition for the Hatta number):
ables of all experiments is included for comparison. A
number of experiments have been performed with insuffi-  2
1 FGL
cient tracing in part of the piping from the reactor vessel k¼  (13)
to the condenser causing a partial condensation of the D VR  a  cA,i
vapour and subsequent reflux of components to the reactor.
The rate of oxygen transferred to the liquid phase (FGL) For w , 0.3 the bulk concentration can be calculated
was calculated from the difference between the gas feed from a balance for oxygen in the liquid phase:
and off-gas flow. Both the liquid and the gas phase were
assumed to be ideally mixed; the residence times for the 0 ¼ fV  cA,b  ð1  1Þ  VR  k  cA,b þ J  VR  a (14)
liquid were in the order of one hour. To address the gas
phase mixing qualitatively a visual check was performed The outflow of dissolved oxygen is considered to be much
in an glass vessel comprising identical baffle and stirrer smaller than the reaction term and the transfer rate through
configuration. With cyclohexane as liquid at room tempera- the gas – liquid interface; therefore equation (14) can be
ture and nitrogen as gas feed, it was observed that at 1600 simplified to:
rpm stirrer speed a significant amount of gas backmixing
occured into the liquid by means of vortex formation 0 ¼ (1  1)  VR  k  cA,b þ J  VR  a (15)
around the stirrer and turbulences behind the baffles. In
addition, it was assumed that no oxygen is present in the with the substitution for J [equation (8)]:
liquid feed and the system operates at steady state. The cA,b 1
oxygen concentration in the liquid at saturation was calcu- ¼ (16)
lated according to Henry’s law. The Henry coefficients cA,i 1 þ Al  w2
were calculated from gas –liquid equilibirum data retrieved where Al is the Hinterland ratio:
from the Dortmund Data Bank (program version DDBSP
2003; http://www.ddbst.com, president and CEO of (1  1)  kL
DDBST is Prof. Dr J. Gmehling, access date 26 November Al ¼ (17)
aD
2004). Typical values are in the range of 1500 – 2000 bar.
The partial pressure of oxygen was determined from the Combining the expression for the bulk concentration
off-gas and condensate composition. In Figure 3 the con- with the flux and the observed conversion rate gives an
version rates for oxygen are shown as function of the equation that can be solved with respect to the kinetic
oxygen concentration in the liquid. The conversion rate rate constant:
for oxygen is defined as the transfer rate of oxygen divided  
cA,i
by the liquid volume; in this way the outflow of dissolved FGL ¼ VR  a  kL  cA,i  (18)
1 þ Al  w2
Solving for k gives:
kL2 FGL
k¼  (19)
Al  D VR  kL  a  cA,i  FGL
Values for specific surface areas were estimated with the
correlation of Shridhar and Potter (1980), diffusion co-
efficients were estimated with the correlation of Wilke
and Chang (1955). Mass transfer coefficients were esti-
mated for water with the equation of van’t Riet (1979).
To translate the mass transfer coefficients from water to
toluene at reactor conditions, a result from Higbie’s pen-
etration theory was applied:
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
kL (toluene, 1508 C) D(toluene, 1508 C)
¼ (20)
kL (water, 258 C) D(water, 258 C)

The experimental conditions in the experiments were

such that the volumetric mass transfer coefficient could
Figure 2. Experimental set-up. be taken as constant at a value of 0.75 s21. The interfacial

Trans IChemE, Part A, Chemical Engineering Research and Design, 2005, 83(A2): 187–195
190 HOORN et al.

Table 1. Data from a series of standard experiments.

Range of operation including Standard deviation

all experiments in reference experiments

Reactor temperature 8C 140–160 0.8%

Reactor pressure bar 4–7 0.2%
Off-gas inert flow Nl/h 100–300 1.1%
Feed flow toluene kg/h 1–2 1.0%
Feed flow air Nl/h 100–300 0.3%
Flow liquid discharge kg/h 0.5–1.5 2.1%
Flow condensate kg/h 0.5–1.5 2.3%
Off-gas oxygen level vol% 0–6 32.5%
Condensate Balc mol/h 0.0003–0.005 12.5%
Bald mol/h 0.003–0.02 7.3%
BzA mol/h 0–0.02 20.8%
H2 O mol/h 0.6–1.6 11.0%
Tol mol/h 2.1–5.6 4.8%
Liquid discharge Balc mol/h 0.003–0.05 3.9%
Bald mol/h 0.03–0.2 1.8%
BzA mol/h 0.4–1.1 3.3%
Tol mol/h 1.5–10.9 1.5%

area was taken as a constant at 300 m2/m3. Under the
hypothesis that the reaction is fast, i.e., w . 2, the
calculated rate constant [equation (13)] is 100 s21
(+56 s21, standard deviation for the experimental obser-
vations). The corresponding Hatta number on basis of this
rate constant can then be estimated to be 0.60 + 0.16.
This is not in line with the assumption of a fast reaction
with w . 2. For the hypothesis of a slow reaction, i.e., w
, 0.3, the calculated rate constant is 2.2 + 3.2 s21
giving a Hatta number of 0.079 + 0.048 which is in line
with the assumption made for the Hatta number. Therefore,
it is concluded that on basis of the experimental obser-
vations the reaction rate is slow in comparison to the
mass transfer rate of oxygen.
MODELLING OF THE EXPERIMENTAL REACTOR
A model was constructed to determine the kinetic par-
ameters under conditions of mass transfer limitations
while taking into account partial condensation of the
vapour phase. To include a description of the partial con-
densation between the reactor and the condenser the
model comprised two flash calculations for the reactor
part of the set-up. A scheme of the model is given in
Figure 4. For an overview on the model equations see
Appendix. The model calculations were performed in
Aspen Custom Modeler. Thermodynamic properties for
the liquid phase were calculated with Aspen Plus Proper-
ties. The pure component properties were taken from the
DIPPR Handbook (http://dippr.byu.edu, American Insti-
tute of Chemical Engineers, access date 25 May 2004) or
DDBST (http://www.ddbst.com). The interaction between
components was described by the NRTL-model. The
thermodynamic property files generated in Aspen Plus
were included in the ACM file. The liquid from the
second flash is totally refluxed to the first flash (‘liquid
recycle’). The temperature of the second flash was varied
for each experiment in such a way that the differences
between experimental and model values for the gas and
liquid mole flows of all components were minimized. The
calculations to determine the temperatures of the second
flash were performed with a simulated combined feed of
the liquid discharge and vapour flows. For experiments per-
formed with the correct tracing, the temperature of
the second flash was fixed at the temperature of the first

Figure 3. Observed conversion rates for oxygen as function of the oxygen Figure 4. Schematic drawing of model. The symbols in this scheme are
concentration in the liquid at saturation. utilized in the overview of model equations.

Trans IChemE, Part A, Chemical Engineering Research and Design, 2005, 83(A2): 187–195
 MODELLING TOLUENE OXIDATION 191

Table 2. Statistics for vapour–liquid calculations.

SSR without reflux Standard deviation

SSO (mol2/h2) SSR (mol2/h2) Standard deviation (mol2/h2) without reflux

Condensate Balc 3.24 . 1025 5.61 . 1026 35.5% 8.74 . 1025 127.5%
Bald 9.29 . 1024 1.69 . 1024 21.7% 1.38 . 1023 61.3%
BzA 4.45 . 1024 1.87 . 1024 34.2% 1.34 . 1024 37.3%
H2O 1.26 . 101 2.23 . 1021 11.0% 2.57 . 1021 12.6%
Tol 1.61 . 102 2.72 . 100 11.0% 3.34 . 101 36.1%
Liquid discharge Balc 6.92 . 1023 5.61 . 1026 2.4% 8.74 . 1025 9.4%
Bald 1.53 . 1021 1.69 . 1024 3.0% 1.38 . 1023 8.7%
BzA 5.09 . 100 1.87 . 1024 0.5% 1.34 . 1024 0.4%
Tol 4.84 . 102 2.72 . 100 11.0% 3.34 . 101 38.5%

flash. The temperatures of the second flash were 5 to 158C
lower than the reactor temperature depending on the
composition of the condensate flow. The liquid recycle
amounts to 10 – 20% (mole flow basis) compared to the
vapour flow feeding the second flash. The statistical results
of the calculations are given in Table 2. The kinetic
parameters in Figure 1 were estimated by a least squares
minimization procedure. In the estimation procedure (and
in the subsequent model calculations) the temperatures in
the second flash were fixed at the values calculated earlier.
The most suitable variables as experimental data points and
the optimal weighting were identified after some trial and
error. Best results were obtained with the toluene conver-
sion, selectivities for benzyl alcohol, benzaldehyde and
benzoic acid and the oxygen off-gas concentration. The
weights of the experimental values for a particular variable
were set as reciprocals of the average values, except for
oxygen which was given a lower value (10% of the recipro-
cal average). The results are given in Table 3. In Table 4
the sums of the squared residuals of model prediction
and experimental values (SSR) are shown. The model
results for the data applied in the estimation procedure
are shown in the parity plots in Figures 5 –8. Comparison
between experimental values and model calculations for
the oxygen concentration in the off-gas are shown in
Figure 9. The residuals for benzyl alcohol, benzaldehyde
and benzoic acid selectivity are shown in Figures 10
and 11. Residual values are given as the difference in
selectivity between experiment and model relative to the
observed experimental value. The calculated pseudo
first-order rate constant is kfo ¼ 295 + 14 s21, the Hatta
number is w ¼ 0.92+0.05 and the enhancement factor is
E ¼ 1.36+0.03.
DISCUSSION
From Table 1 it appears that the input process parameters
are controlled with acceptable accuracy of approximately
1%. The response variables differ in accuracy in the
range of 1.5% for toluene in the liquid discharge flow to
20% for benzoic acid in the condensate flow. Benzoic
acid concentrations in the condensate flow are difficult
to analyse because at room temperature the condensate sep-
arates into a water and an organic phase. The low accuracy
for the oxygen measurement is a combination of the
analytical measurement and the length of the sample pipe
from the off-gas vessel to the instrument. The simulation
of the partial condensation in the piping to the condenser
vessel is successfully performed by modelling the reactor
part with an additional flash calculation (Figure 4) and
this is illustrated when the results in Table 2 are compared.
The differences between model predictions and experimen-
tal observations are clearly reduced with respect to the situ-
ation where there is no liquid recycle. When compared
to the indication of experimental errors in Table 1, the
model describing the partial condensation has a scatter in
the model predictions that is much closer to the

Table 3. Estimated parameter values and correlation matrix.

Parameter values Reactions Correlation matrix

3
k12 0.015 + 0.004 m /mol/s Tol ! Balc k12 k3 k4
Tol ! Bald k12 1
k3 1.35 + 0.36 m3/mol/s Bald ! BzA k3 0.98 1
k4 3.9 + 1.0 m3/mol/s Balc ! Bald k4 0.99 0.98 1

Table 4. Sum of squares for experiments and model.

Experiment SSO Model SSR

1
O2 off-gas 7.2 . 10 4.3 . 101
Tol conversion 7.2 . 1022 1.8 . 1023
Balc selectivity 9.6 . 1023 1.4 . 1024
Bald selectivity 2.4 . 1021 1.7 . 1023
BzA selectivity 7.5 . 100 1.6 . 1023 Figure 5. Comparison between experimental values and model
calculations for the conversion of toluene.

Trans IChemE, Part A, Chemical Engineering Research and Design, 2005, 83(A2): 187–195
192 HOORN et al.

Figure 6. Comparison between experimental values and model Figure 9. Comparison between experimental values and model calcu-
calculations for the selectivity of benzyl alcohol. lations for the oxygen concentration in the off-gas.

Figure 7. Comparison between experimental values and model

calculations for the selectivity of benzaldehyde.

Figure 10. Residuals between experimental and model values as function

Figure 8. Comparison between experimental values and model of the residence time based on the liquid feed flow rate; (W) ¼ benzyl
calculations for the selectivity of benzoic acid. alcohol; (A) ¼ benzaldehyde; (þ) ¼ benzoic acid.

Trans IChemE, Part A, Chemical Engineering Research and Design, 2005, 83(A2): 187–195
 MODELLING TOLUENE OXIDATION 193

determination of the reaction rate order in oxygen the

results in Figure 3 do not exclude other than simple first
order kinetics. The choice for a first-order oxygen depen-
dence is quite arbitrary and has been done for practical
reasons, a zero-order model is more difficult to handle
because of the discontinuities with the reaction rates.
When some reaction rate equation other than a power
law is required to improve the model accuracy, mass trans-
fer rates have to be solved by numerical procedures. The
combination of these numerical solutions with parameter
estimation procedures has not been performed so far. The
value of the Hatta number that was directly derived from
the experimental observations indicates that the oxidation
of toluene can be considered to be slow. The Hatta
number of 0.09 derived from the experimental observations
is in good agreement with the value calculated with the
model. The enhancement factor calculated by the model
is also close to unity. The results obtained with the
model show a more narrow distribution of errors. The stan-
dard deviations in the average Hatta number is smaller (5%
relative) than with the experimentally determined Hatta
number (50%). The low values of Hatta and the enhance-
ment factor imply the presence of an oxygen concentration
in the liquid bulk, suggesting that reaction of oxygen with
some kind of radical intermediate can not be the rate-deter-
mining step in the overall mechanism. This is in general
agreement with the picture for autoxidation reactions
(Helfferich, 2001).

CONCLUSIONS
Experiments performed in a reactor that was operated
under industrial conditions have been successfully applied
in the estimation of kinetic parameters in a model with
inclusion of the description of mass transfer phenomena.
Analyses of the mass transfer characteristics reveal that
toluene oxidation under industrial conditions is a slow reac-
tion with respect to mass transfer.

Figure 11. Residuals between experimental and model values as function

of the experimental observed oxygen concentration in the off-gas;
(W) ¼ benzyl alcohol; (A) ¼ benzaldehyde; (þ) ¼ benzoic acid. NOMENCLATURE
a interfacial area per unit volume of dispersion, m2/m3
Al Hinterland ratio
experimental error than the model version without the c concentration, kmol/m3
reflux mechanism. Tables 3 and 4 summarize results of cA,b concentration of A in the liquid bulk, kmol/m3
cA,i concentration of A at the gas–liquid interface, kmol/m3
the estimated kinetic parameters. The values of the standard cO2,sat saturation concentration of oxygen, kmol/m3
deviations as well as the correlation parameters are in cO2,out oxygen concentration in the gas outflow, kmol/m3
acceptable limits for further scale-up purposes, but D diffusion coefficient, m2/s
improvement is still desired. The residual sums are accep- E enhancement factor
F flow, kmol/s
table in comparison to the experimental values. This picture FGL flow across gas–liquid interface, kmol/s
is confirmed by the graphs of the residual plots of the J flow across gas–liquid interface, kmol/m2/s
toluene conversion and the selectivity’s of benzyl alcohol, k reaction rate constant (second order), m3/kmol/s
benzaldehyde and benzoic acid (Figures 5 – 8). The parity k reaction rate constant (first order), s21
plot for oxygen exhibits a fair off-set (Figure 9). This is kfo reaction rate constant (pseudo first order), s21
kL liquid side mass transfer coefficient, m/s
caused by the low weighting value for oxygen which was kL a volumetric liquid side mass transfer coefficient, l/s
based on the inaccuracy of the experimental value. The n reaction order
residuals are larger when the liquid residence time p partial pressure, bar
increases or when the oxygen concentration is higher. Ptot total pressure (absolute), bar
R reaction rate, kmol/m3/s
(There is no direct relation between liquid residence time VR dispersion volume (liquid þ gas), m3
and oxygen concentration because the gas flow rate is set w mole fraction in liquid, mol/mol
independent of the liquid feed flow rate.) For the x mole fraction in liquid, mol/mol

Trans IChemE, Part A, Chemical Engineering Research and Design, 2005, 83(A2): 187–195
194 HOORN et al.

xO2,sat oxygen mole fraction in liquid at saturation, mol/mol Scott, E.J.Y., Chester, A.W., 1972, Kinetics of the cobalt-
y mole fraction in vapour, mol/mol catalyzed autoxidation of toluene in acetic acid. J Phys Chem, 76:
z mole fraction in vapour, mol/mol 1520–1524.
Shridhar, T., Potter, O.E., 1980, Interfacial areas in gas-liquid stirred
Greek symbols vessels. Chem Eng Sci, 35: 683–695.
1 gas volume per unit volume dispersion, m3/m3 Suresh, A.K., Sridhar, T., Potter, O.E., 1988, Autocatalytic oxidation
x conversion, mol/mol of cyclohexane—mass transfer and chemical reaction. AIChE J, 34:
f fugacity coefficient 81–93.
fV volumetric flow, m3/s Westerterp, K.R., van Swaaij, W.P.M., Beenackers, A.A.C.M., 1984,
w Hatta number Chemical reactor design and operation, 371–377. (John Wiley &
rmol,L liquid molar density, kmol/m3 Sons, New York, USA).
s selectivity, mol/mol Wilke, C.R., Chang, P., 1955, Correlation of diffusion coefficients in dilute
tL residence time, s solutions. AIChE J, 1: 264–270.

Sub/superscripts (see also Figure 4 for descriptions) The manuscript was received 2 June 2004 and accepted for publication
i any component after revision 13 December 2004.
BALC benzyl alcohol
BALD benzaldehyde
BzA benzoic acid
H2O water
N2 nitrogen APPENDIX
O2 oxygen
TOL toluene Overall pseudo first-order rate constant for oxygen
F feed flow to Flash 1
G gas vapour flow from Flash 2 3 1
L liquid discharge flow from Flash 1 k fo ¼  k12  cTOL þ  ðk3  cBALD þ k4  cBALC Þ
V vapour flow from Flash 1 to Flash 2 2 2
R liquid recycle flow from Flash 2 to Flash 1
Definition of Hatta number and mass transfer equations
Abbreviations pffiffiffiffiffiffiffiffiffiffiffiffiffi
ACM Aspen Custom Modeler kfo  D
SSO Sum of Squared Observations w¼
SSR Sum of Squared Residuals
kL
FGL ¼ J  VR  a
 
REFERENCES pffiffiffiffiffiffiffiffiffiffiffiffiffiffi cO2,sat  (cO2 )=(1 þ w2 )
E ¼ 1 þ w2 
Bateman, L., 1951, Olefin oxidation. Quart Revs (London), 8: 147–167. cO2,sat  cO2
Bhattacharya, D., Guha, D.K., Roy, A.N., 1973, Liquid phase air oxidation  
of toluene to benzoic acid. II. Kinetics and mechanism. Chem Age India, J ¼ kL  cO2,sat  cO2  E
24: 87–90.
Borgaonkar, H.V., Raverkar, S.R., Chandalia, S.B., 1984, Liquid phase Reaction rate expressions (according to Figure 1)
oxidation of toluene to benzaldehyde by air. Ind Eng Chem Prod Res
Dev, 23: 455–458.
Cao, G., Servida, A., Pisu, M., Morbidelli, M., 1994, Kinetics of p-xylene R1 ¼ k12  cTOL  cO2
liquid-phase catalytic oxidation. AIChE J, 40: 1156–1166. R2 ¼ k12  cTOL  cO2
Helfferich, F.G., 2001, Kinetics of homogeneous multistep reactions, in
Compton, R.G. and Hancock, G. (eds). Comprehensive Chemical R3 ¼ k3  cBALD  cO2
Kinetics, vol. 38, 283–286 (Elsevier, Amsterdam, The Netherlands).
Hikita, H., Asai, S., 1964, Gas absorption with (m,n)-th order irreversible R4 ¼ k4  cBALC  cO2
chemical reaction. Int Chem Eng, 4: 332 –340. RBALC ¼ R2  R4
Keading, W.W., Lindblom, R.O., Temple, R.G., Mahon, H.I., 1965,
Oxidation of toluene and other alkylated aromatic hydrocarbons to ben- RBALD ¼ R1  R3 þ R4
zoic acids and phenols. Ind Eng Chem Process Des Dev, 4: 97–101.
Kondratiev, V.N, 1969, The theory of kinetics, in Bamford, C.H. and RBzA ¼ R3
Tipper, C.F.H. (eds). Comprehensive Chemical Kinetics, vol. 2, RH2O ¼ R1 þ R4
165– 173 (Elsevier, Amsterdam, The Netherlands).
Lozar, J., Falgayrac, G., Savall, A., 2001, Kinetics of the electrochemically 1
assisted autoxidation of toluene in acetic acid. Ind Eng Chem Res, 40: RO2 ¼ R1   (R2 þ R3 þ R4 )
6055–6062. 2
Morimoto, T., Ogata, Y., 1967, Kinetics of the autoxidation of toluene RTOL ¼ R1  R2
catalyzed by cobaltic acetate. J Chem Soc (B), 62– 66.
Mulkay, P., Rouchaud, J., 1967, Solvatation des catalyseurs d’oxydation
en phase liquide homogène. Bull Soc Chim Fr, 12: 4653– 4657. Definition for toluene conversion
Pohorecki, R., Baldyga, J., Moniuk, W., Podgórska, W., Zdrójkowski, A.,
Wierzchowski, P.T., 2001, Kinetic model of cyclohexane oxidation.  
Chem Eng Sci, 56: 1285–1291.
FF,TOL  FL,TOL þ FG,TOL
x¼
Quiroga, O.D., Gottifredi, J.C., Capretto de Castillo, M.E., 1980, Liquid FF,TOL
phase catalytic toluene oxidation, formulation of a kinetic model.
Lat Am J Chem Eng Appl Chem, 10: 77 –88.
Russell, G.A., 1957, Deuterium-isotope effects in the autoxidation of
Definition for benzoic acid, benzaldehyde and benzyl
aralkyl hydrocarbons. Mechanism of the interaction of peroxy radicals. alcohol selectivity
J Am. Chem Soc, 79: 3871–3877.
van’t Riet, K., 1979, Review of measuring methods and results in nonvis- FL,i þ FG,i
cous gas-liquid mass transfer in stirred vessels. Ind Eng Chem Process si ¼ P P
Des Dev, 18: 357 –364. j FL,j þ j FG,j

Trans IChemE, Part A, Chemical Engineering Research and Design, 2005, 83(A2): 187–195
 MODELLING TOLUENE OXIDATION 195

Definition for off-gas oxygen concentration Mass balances for all components except oxygen for
Flash 1 and Flash 2 (see Figure 4)
FG,O2
cO2,out ¼
FG,N2 þ FG,O2 fL,i  xi ¼ fV,i  zi
fG,i  yi ¼ fR,i  wi
Mole fractions and concentrations for all components
FF,i þ FR,i þ V  Ri ¼ FL,i þ FV,i
ci ¼ rmol,L  xi FV,i ¼ FG,i þ FR,i
FL,i
xi ¼ P Mass balances for oxygen
i FL,i
FG,i fL,O2  xO2,sat ¼ fV,O2  zO2
yi ¼ P
i FG,i fG,O2  yO2 ¼ fR,O2  wO2
FVAP,i FF,O2 ¼ FGL,O2 þ FV,O2
zi ¼ P
i FVAP,i FGL,O2 þ FR,O2 þ VR  RO2 ¼ FL,O2
FREC,i FV,O2 ¼ FG,O2 þ FR,O2
wi ¼ P
i FREC,i

Trans IChemE, Part A, Chemical Engineering Research and Design, 2005, 83(A2): 187–195